Hot take after 4 years of homelabbing: the arrs and qbit aren't a homelab, they're a media server.
The unlock is self-hosting your own tools. Mine now transcribes my Arabic lessons and pipes new vocab straight into a self-hosted Anki clone.
That's one of 6 tools I've built that I actually use daily.
@beer4brekkie try https://t.co/FQDx0F8luc
it will auto import all your twitter bookmarks on first sync, and then keeps them in sync
then you use text search or ai chat or claude mcp to find what you need
@wh0sumit try https://t.co/FQDx0F8luc and let it handle saving + parsing + retrieval.
use obsidian.md for persistence
use claude code as your orchestrator
I used to feel guilty spending time on social media.
not anymore. not because I got more disciplined.
because everything I save goes into @animus_so and comes back the moment I need it.
scrolling became research.
1600 PRs to my side project. solo founder.
the cheat code: previews at every layer.
Railway preview -> Vercel preview -> GCloud preview
auto cleanup after merge
you stop being precious about merging when every PR is a throwaway env.
Updated the landing page for https://t.co/FQDx0F8luc again.
Getting a consistent stream of page views but not enough trial sign ups
Don't think the landing page was communicating the value prop clearly
Your first startup should take 1 year to build.
Your second, 7 months.
Your third 3, months.
Your fourth, fifth and sixth should be live within 2 weeks.
If you skip this, you're not ready for what will hit you and you'll miss out on the best part of being a solo founder.
The lessons from the first 3 failures are what define everything else you will do
We're optimising for this at https://t.co/mzzY07e6FG
It can save anything from anywhere with one click (insta, twitter, youtube, tiktok, websites, articles, pdfs)
It understands audio, video, text or a combination of all three
It auto organizes into collections at the click of a button.
The only thing left for the human is to recall through built in chat or mcp. So when you're looking to brush up on a topic, you have everything you need at your fingertips from across your library.
Has anyone solved scene detection by a reliable measure?
I have transcription, I have ocr. But sometimes the thing is in a frame on a video. Not in the audio.